Yahoo Finance

Chris Siu

Thank you, Gayn. Before reviewing our financial results, I would like to provide an update on the integration of our InCal acquisition, which closed last July. Our plan to consolidate personnel and manufacturing into Aehr's Fremont facility is progressing well, and we remain on track to complete the integration by the end of this fiscal year on May 30.
As part of this effort, we have upgraded water and power systems and added new clean rooms to our remodel Fremont headquarters to ensure the infrastructure supports the needs of both organizations. We'll be shutting down the InCall facility no later than the first quarter of fiscal 2026. Since the acquisition, Aehr has committed significant financial and human resources to successfully integrate InCal into our operations. We have completed the migration of InCal's financial, HR and manufacturing functions into Aehr systems.
In addition, we have finalized the transfer and documentation of our product designs, source code, and work instructions for assembly and test into Aehr release processes. I want to extend my sincere thanks to both teams for their dedication and outstanding execution throughout this integration.
Turning to our Q3 performance, which included a full quarter on the financial results from the InCal acquisition. Our Q3 results exceeded the analyst consensus on both the top and bottom lines. While we faced a challenging environment marked by continued softness in the silicon carbide power semiconductor market. We were encouraged by our success in penetrating the artificial intelligence market.
With AI processors burn-in now representing over 35% of our business this year. During the third quarter, we had four customers representing over 10% of total revenue, and two of these customers are new customers that target the AI market. Revenue for the third quarter totaled $18.3 million, a 142% increase compared to the $7.6 million in Q3 last year. The significant year-over-year revenue growth was driven primarily by the shipments of our new high-power FOX-XP solution. for wafer-level production test and burning of AI processors delivered to our first AI processor customer, which we announced in December.
In addition, we're pleased with the significant progress we've made integrating products from our acquisition of Inkal into our product portfolio to address the AI market opportunities. System sales from our Sonoma, Tahoe, and Echo package part burn-in products made a strong contribution to our third-quarter revenue. We believe our strategy to broaden Aehr's offerings and diversify behind silicon carbide applications is beginning to show positive results, both operationally and financially. WaferPak revenues were $5.9 million, accounting for 32% of our total revenue in the third quarter, a decrease from 63% in the same period last year.
The company recognized bookings of $24.1 million in the third quarter of fiscal 2025 compared to $9.2 million in the second quarter fiscal 2025. Our backlog at the end of the quarter was $18.2 million. Since the end of the third quarter of fiscal 2025, we have received $3.6 million in additional bookings. With these recent bookings, our effective backlog now stands at $21.8 million.
Non-GAAP gross margin for the third quarter was 42.7% compared to 42.5% in the same period last year. The overall change in gross margin was flat, primarily due to a much higher overall revenue level than in Q3 last year, offset by a less favorable product mix and onetime items.
Even though our revenue in the third quarter was much higher than that in the same period last year, our non-GAAP gross margin was lower than expected due to certain onetime charges to our cost of revenue. After we upgraded our ERP from our legacy system to Oracle NetSuite, our new ERP allowed us to account for standard costs in inventory more precisely and revised certain accounting estimates, which resulted in a onetime charge in the third quarter.
Additionally, we incurred high manufacturing overhead due to underabsorption as we had lower utilization of our manufacturing capacity due to the renovation of our Fremont manufacturing Facilities. Non-GAAP operating expenses in the third quarter were $6.3 million, reflecting a 34% increase from the $4.7 million in Q3 last year. This year-over-year rise is primarily attributed to the inclusion of InCall's expenses in our financial results, along with higher legal and professional service fees.
We anticipate incurring additional legal expenses in the upcoming quarters as we strive to protect our intellectual property rights in China and defend against the class action and derivative complaints in the United States, which we believe like merits.
Non-GAAP net income for the third quarter, excluding the impact of stock-based compensation, acquisition-related costs, the fair value adjustment to inventory related to the acquisition, amortization of intangible assets and the accelerated benefits paid to an executive who passed away unexpectedly in December was $2.0 million or $0.07 per diluted share. This compares to a non-GAAP net loss of $888,000 or negative $0.03 per diluted share in the third quarter of fiscal 2024.
Turning to our balance sheet. At the end of Q3, our cash, cash equivalents and restricted cash totaled $31.4 million, down from $35.2 million at the end of Q2. During the quarter, we used $1.6 million in operating cash flows, primarily to pay our supplies and service providers. We have no debt and continue to invest excess cash in money market funds.
In the third quarter, we earned $270,000 in interest income. As Gayn mentioned, considering the secondary effects of the tariff announcements on our recent and potential new customers, along with the uncertainty this quarter regarding possible pauses or delays in customer orders, shipments or supply chain delivery delays. We are temporarily withdrawing our guidance for our current fiscal 2025 year ending May 30.
We'll reassess our guidance policy as clarity develops. Looking ahead to fiscal 2026. We are encouraged by the growing number of engagements with both current and potential customers. as well as the long-term growth potential across our diverse target markets. Our strategic expansion into high-growth sectors, including artificial intelligence processors, gallium nitride powered semiconductors.
Data storage devices, Scan Photonics integrated circuits, and FLASH memory opens new opportunities to attract customers and drive revenue growth.
Lastly, looking at the Investor Relations calendar. Aehr will be participating in two investor conferences over the next couple of months. We will be meeting with investors at at the Craigh-Hallum Institutional Investor Conference taking place in Minneapolis on May 28. And we will be presenting and meeting with investors on June 3 at the William Blair 45th Annual Growth Conference taking place in Chicago. We hope to see some of you at these conferences.
This concludes our prepared remarks. We're now ready to take your questions. Operator, please go ahead.

Question and Answer Session

Operator

(Operator Instructions) Christian Schwab, Craig-Hallum.

Tyler Burmeister

This is Tyler on behalf of Christian. So maybe first to start on the tariffs and the uncertainty here. I guess, any color on maybe which end markets you are seeing maybe the most potential impact or most potential uncertainty being caused on the near term here.

Gayn Erickson

I think we certainly prepared lots of different things. That's a good question. No, it's not market. Sometimes it's more customers and geographies, right? So in this case, I'll give you one example.
Our hard disk drive customer has systems that are going to be going to a location outside the US. It includes a prober, heart. We're basically worked with a supplier that developed a high-power prober specific for this requirement. Power companies are out of Japan or Korea, basically, okay, without saying which one it was, okay? They both are currently subject to a tariff.
If that prober landed in the port tomorrow, it's going to get hit by a tariff. If it lands Friday, it may not. And so I think people are like, well, let's wait and see what happens. I mean because I don't think anybody believes that if you write a check for a tariff, you get it back necessarily.
Now having said that, we're also looking at tariff drawbacks, which is basically a process by which historically with duties, you can. If we import something do assembly and then export it out, you can actually recover that up to 99% of it or so, although there are some people saying that the people that help you with that get a cut. So we're working through that as well.
Now in the meantime, I'm going to drop ship the probers directly to that customer past the first one. So the very first one is being impacted by, which is kind of the over under like with respect to the end of our fiscal year because we're going to do some integration of it here. We may have to relook at that.
So it's that sort of tactical thing. And given a couple of days, we're what, three days in, where Chris and I were saying, could we have another week, let things settle out before this call. Again, a lot of the material we have, we have already on hand. It will never be subject to any tariffs, I guess, unless we ship it out to someone who then increases their tariffs.
Last night, I went to bed thinking that Europe was going to come back and raise their tariffs. Today, they wake up and they say zero for zero. So it's kind of hard to look at it. And honestly, we're not I don't think anyone is so desperate to try and solve these things in the next couple of weeks, either if you look at just schedules. And so I think we can manage through this.
And then even then we start looking at material in many cases because remember, our inbound material tariffs are on the material, we then, of course, have a markup or margin to ship it. And so a tax on the material of 10% is not a tax of the price. That makes sense because it's because -- so we kind of walked through that. I would say that there's not a specific market.
Obviously, China faces own challenge in that I think they came back and said it's 104% starting tomorrow. We actually have, if you specifically look at kind of risks related to China, just a couple of things.
Aehr has basically had no revenue shipments into China for the last few years. That includes both our wafer level stuff and the inCal products. So we're not depending upon shipping a lot of revenue into China. That's not been the plan.
And so if China comes back and says, we're going to halt shipments or we're going to double the tariffs or whatever, it doesn't really affect us right now, okay? Similarly, we have been pretty happy and proud that we almost do nothing in China from a supply chain perspective. I said that in previous calls, we actually have some things like blank printed circuit board fabs, but not the assembled ones.
And so it's pretty simple, and we have secondary suppliers for that just from a risk perspective. we continue to buy from them because the landing cost made more sense, but we also can supply it on several other countries. We wake up in like which country should we supply from. It's like give it a day. So there's a little bit of like settling out on this.
And that, again, isn't even so much of the cost implications to margins as potential supply chain risks, what if they put that in a crate, it's shipped over here and it gets stuck in customs because people are trying to figure out what country it came from and what data they come in because that's the tariff that's applied to.
So we're actually doing things to try and avoid anything coming into US for customs in the near term. Not even so much for tariffs just simply don't cluster have it get cluttered up and then you have it in the wrong place. We can redirect things. We have things -- we have supply from material and subsystem from all over the world.
They can drop ship directly into a customer internationally. We don't even have to worry about any of that stuff. So we have the infrastructure, we have the ERP systems. We have the legal entities that we can do that. And candidly, we're kind of proud of that supply chain customer -- I mean with the China COVID thing, I think everybody had to take a look at it. We actually were also quite immune from anything during COVID.
But this has been an important thing to tell your customers. What happens as such and such happens, we've said, you're not going to miss a shipment, we're going to be okay, right? So it's kind of an important thing to note. So people understand it.
But at the same time, we can't answer everything. Remember, May 30 is what, seven weeks away. I mean, we're like -- and as you know, plus or minus one system on ours can move the bar. So that's why we just pulled the near-term guidance.

Tyler Burmeister

All right. I appreciate that. That's a bunch of really great color. So I guess following up on that, then giving those comments uncertainty around additional orders and timing of when customers might demand shipments, you had $18 million in backlog in the quarter, $22 million effective since then.
Would it be reasonable or area on the conservative side, at least a reasonable to think that you probably won't even ship all of the backlog you currently have on hand in the quarter? Or is that not a great way to think about it?

Gayn Erickson

One of the keys is if we're going to say we're not going to give guidance, we can't really give guidance. But nevertheless, let me still answer that directly. For sure, some of that backlog is not going to ship this quarter because it's multiple systems for the hard disk drive guy.
And we're only planning to maybe ship one of those. We said that last quarter. So the over under right now is one, maybe a second one for this quarter. but not all of them. So some of those are already planned for that.
The interesting thing is we have multiple customer forecasts for things to ship this month or this quarter that they haven't ordered yet. You're like, okay, is that going to happen? Would that happen that I can still turn that? Some of it, we have material on hand and some of it is stuff that's inbound.
So I'd love to tell you it's that simple. And if it was, I'd probably just give you a new guidance, but it's sort of there's kind of upside and downside to all of this. So I apologize for not being really clear. I think the important thing is we missed it. We missed it by a little whatever. It's going to be close.
It might chip on the next week. That's the thing that's really can be very scary as a public company, right? If you're private, I wouldn't even worry about it. I'm only worried about the customers getting their shipments, getting to them on time. But plus or minus one day on May 30 is a big deal to us as a public company.
So this is more -- the things I'm talking about here, this is just timing stuff, we think, okay? This is more timing things. I don't think the customers that are already buying, they've taken, in some cases, six months to a year to qualify. I mean if you look at the GaN customer, the hard disk drive customer, this AI customer, the Package part burn-in AI customer, we've worked with them for over a year.
They're fully qualified. They're committed. Now the question is how do we make sure that they can get the equipment on time with the least duties or tariffs at all possible?
If you think of another one, Tyler, you can always come back to. Go ahead, what's the next question, sorry?

Tyler Burmeister

No, let's move on past the tariffs. So as we think about next year in growth, you've diversified your revenue nicely away from silicon carbide maybe at least rank order. It sounds like silicon carbide is expected to continue to recover, we would expect to see growth there.
But maybe AI is going to be the primary growth driver as we look out the next year or two. Just any color kind of rank ordering and where you expect to see the growth from your more diversified end markets going forward.

Gayn Erickson

Yes. I mean we're feeling really good about next year. I mean there definitely is evidence that silicon carbide will kind of get back to capital equipment growth and putting more capacity in place.
We're also seeing a shift from low-voltage gate to high-voltage testing. Something we've done by working with several of the OEMs, the actual EV suppliers for quality that they're driving to their customers, which drives them towards us that even if they have equipment, they may need to upgrade it.
That also turns into WaferPack revenues. So we think we can see revenue from our current installed base in both wafer packs and system upgrades, but then also incremental capacity. And there's still a few customers that effectively selected us, if you will, and haven't bought their first systems because they've delayed some of their fabs that Yale is saying they will start to turn on in '26 and '26 is when they say it's calendar, my fiscal '26 starts June some of that capacity may be second half, but we feel some of this already in our first half or so our second half of '25.
Our fiscal year has kind of messed people up. Let's see here. GaN, we're pretty excited about that. There's some we have done a large number -- I mean, a good number of designs that have been qualified into automotive industrial applications. They're now just going into production. So we've kind of got our fingers crossed related to that around their capacity and their needs to grow.
And we're talking with several other customers. We've actually heard our silicon photonics customer is talking now about ramping this year. And that's with the very high power. Remember, we upgraded their fleet to the extra high power system, which needs new wafer packs and needs new system upgrades.
That's encouraging because of all the stuff that's finally going on around the silicon photonic side. But I think we think that the package part burn-in qual is going to continue to grow. There's a lot of new devices coming out with higher power that's going to drive demand.
And then burn-in, the production side of it. So the production burden, we're seeing our first customer ramp. It's in a very visible OSAT. And we're literally shipping all the systems. We finished it this quarter and then they start their first ramp. We think they'll need more systems.
We're also working on some enhancements. I don't want to get ahead of myself here in too much public, but we're working on some automation and some very specific things in the capability to add to that package part burn-in and have been communicating that to the customers and the OSATs right now. And I'm very hopeful and encouraged by the feedback that could drive production -- more and more production burn-in to our systems with full automation with the Sanoma line.
And then lastly, but honestly, maybe most impactful is what's going on with the wafer level. That first customer is expected to ramp, and we already have inbounds from other people. The people are really leaning in to understand what do they need to do? What testability modes do they need to add to their wafer, how can they do this?
Because the implications of being able to weed out the failure of that processor before you stick it into a large package with co-packaged memory, optics and other processors is huge. That could be some of the most exciting stuff that's going on.
Flash memory. Actually, we're spending a good -- a significant amount of R&D money this year. We're spending our money. I'd say our as collectively as shareholders. It's a great investment.
We're going to need to invest even more next year in developing the test solution to go in that. Maybe we could get some revenue at the end of the year, but this is really an early 27 is what we're thinking in terms of the timing. That's why, one, I use those TAMs around '27 kind of capture at that point, we would hope to have flash memory revenue.
Wafer level burn-in across more than hopefully several or more customers, packaged part burn-in across several customers, silicon carbide hitting GaN starting to ramp and the hard disk drive just continuing to buy along with silicon photonics.

Operator

Jed Dorsheimer, William Blair.

Jonathan Dorsheimer

So again, I don't want to ask you a bunch of questions you don't want to answer. Just I guess, first one, just on the balance sheet. I'm curious, AR, your accounts receivable jumped up quite a bit. Is that just from InCal? Is there a change in terms?
Could you just talk about what was going on there? Because I saw that positively, inventories actually came down while revenues grew. So I'm just wondering about the receivables seem to jump out. And then I got a follow-up.

Chris Siu

Yes. So we also have an unbilled receivable in our receivable balance that we were able to recognize revenue that cannot be yet. But I think by now, we have sent out the invoice at this point.

Gayn Erickson

AR, it's called shipments turns the last month of the quarter, not atypical of that and yes, I think that was it. I think we may have had a special term with one of the customers with respect to billing at final shipment or something in the first tool, a couple of those things. I wouldn't read too much into it.
Over time, and I want to be sensitive to this, because we have customers listening as well. We do down payments under most circumstances, sometimes with short lead times and things like that. We will work with the customer, billing them 30% and then two weeks later, hitting them with the other one, sometimes isn't most optimum. We try and be balanced on it.
But in general, we want to make sure that customers have skin in the game when they're placing the orders on these systems. And so there's a little of that. But yes, there's no flags, nothing there's nothing there.

Christian Schwab

Okay. And then just on the AI on the processing side of things in burning, I was wondering if you could go through the value proposition and a little bit more granularity. And what needs to be achieved for this, has this gone from R&D to commercial level? Or do you feel like there's still -- even though the size and scale of the orders might be commercial size compared to that of silicon carbide for this for these customers, do you feel like it's progressed that that's it's gone beyond and what are you looking at to get the determination that, okay, we're going to start ramping on this in a durable way.

Gayn Erickson

So, Jed, I think I understand it. Let me try and answer it while repeating the question a little bit along the way. So if the question is, is this sort of like an NPI prototype like a trial run and they haven't really worked that kind of thing, that's not the case. So this customer, and we've alluded to it before, and we have to be somewhat sensitive to what the what's known publicly and what they've said.
And yes, we have not identified who it is yet, okay? But what we have said is that they were doing, which is very similar to a lot of other AI guys they were doing this burn-in at what's known as system level test, okay? So this is very late in the process, basically damn the product side of things, okay?
And the problem with doing at a system level test is that it's very late in the process and any processing implications, et cetera, and you have to use a much more different and sophisticated piece of equipment or the equipment itself to burn-in itself in to actually find these failures.
The other challenge is that in a system level test, you often have this memory or this processor is coexist. And I'm going to be a little careful about too much detail on this because people are all trying to figure out is that what kind of AI processor is it? We hope to come much more public with it. The customers talked about doing that. But it exists with other things, memories, infrastructure, heat sinks, power supplies and all this.
Burn-in, I know a lot of people understand is what it sounds like. You basically are heating it up and you're actually adding more power to it to stress the device to a point that will weed out infant mortality or weak devices. You do not want to break it, okay?
So the trick is to try and get it right to the edge to accelerate this stress test to weed out things as short a time as possible, okay? And generally speaking, the more power you put in and the more heat you put in the faster you can do it.
If you put in too much power and too much heat, you rack the part, right? Now so it turns out each device has a different optimum spot flash memories like 85 degrees, DRAM can be 105 degrees, a microprocessor is 125 degrees, silicon carbide is 140 degrees. They're all at different numbers. If you put them all into the same package, and I'm not saying you put silicon carbide into that package, okay?
But let me throw on silicon photonics. There's actually two burning temperatures for it. You do one process at one temperature and one at the other for ideal. Now stick them all into a large package. What do you burn it in at? It's a problem.
So if you were to put it in an oven and you burned in the processor 125, you could rec the DRAM, okay? You certainly can't bring the processor to 140 degrees like you do the silicon photonics device.
So what you do is you bring it to the lowest common denominator. But here is another problem. That system is not designed to be run hot -- the point is it's being run cold with big heat sinks and water cooling, think of all the stuff that's going on in these racks. So it's normally trying to run at 30 degrees.
And I use Centigrade -- sorry for everybody, a little bit one rooom temperature, right? That's what you are trying to run it at, but when you burn it in, we're running it at 125 degrees centigrade more than boiling water, okay?
You can't do that with a system. You would run everything. You'd all the printed circuit boards, you recall the memory, the power supplies would all die that's not good. So you burn it in way lower temperature, maybe 60 degrees. Well, if you do that, then the burden time is like 10 times longer.
So if you actually move it to the wafer level, I can burn in your memory right within a few degrees or 125 degrees and keep it from going in a thermal runaway. That minimizes the burn-in time. I perfectly optimize that.
And you take that device and now when you put it in the system, it will not fail. It's a big deal, it's a really a big deal. And by the way, like on a packaged part burn-in systems, which we have, as I was one of my favorite things.
I have a front row seat on the Sonoma systems. We're testing these packages. And we're testing the memory stacks in there and the processors and the chipsets, and we have to actually test those at different conditions.
It's a very interesting challenge, but it's way more effective if you can break it down and do it at the memory at the wafer level. But until now, no one has ever even imagined doing that. And we now have the technology with the power. The power supplies the capability, the JTAG and MBIST patterns, and functionality and reading right capability to actually test these things at a wafer level at a cost-effective way that's a real alternative to doing system-level test or a package level burn-in test. And by the way, if you want package level burn in test, I'll sell them to you all day long with my Sonoma systems.

Jonathan Dorsheimer

It does. What's the stability of the power over that duration? That seems like it would be a critical feature in terms of -- within a tight tolerance.

Gayn Erickson

Yeah. Some of the many tricks and the keys are being able to actually regulate and drive power to each device on the wafer individually. That doesn't seem intuitive, right? In traditional package part burn-in systems, like those from our competitors, they have bulk power supplies that supply to the devices, all with the same voltages, they achieve a parallelism by actually trying to share all that.
That didn't work very well because each of the devices need a different voltage Sonoma. One of the greatest things about it as a package part burn-in system is we can individually program the power for every device that's being tested. We'll actually read an ID in the device, come back, adjust our power and tune it for that device for an optimum burn in.
That's an amazing thing that people don't have in packaged part burn-in we do in Sonoma. We're doing the same thing at wafer level, okay? And there's more tricks and I'm just not going to do them publicly. There's other things that we've done.
In fact, there are some patents you can go and find that we've done. We can implement in our wafer level burn-in system in our wafer pack that you just can't do in a normal wafer probing environment either. That allow us to be able to test these devices.
But it's pretty fun talking with the customers because they kind of light up as you talk about what you can do and they're like, wait a minute, I never even thought of that before. I'll give you one other hint, okay?
One of the reasons everyone says this will never work at wafer level burn-in is because these devices might -- one core might take 100 amps, maybe it takes multiple hundred tabs, okay? So what you do is you have a 100 power supply and you run it through a probe card and the probe card needles can only do maybe 0.5 an amp or 50 milliamps I heard somebody say the other day, okay?
And then so I had to put 100 of them in parallel, whatever the math is, sorry, I always tough online like this to get you to 100 amps, okay? Or 1,000 pins to get to 100 amps with all this. What happens at that section of the device fails in short, it will ruin your probe card. So when people -- when we first engage them, they're like, this is absolutely impossible because every wafer, I have a device that shorts in the mill to burn in and it's going to ruin your probe card.
And we have features in our wafer level burn-in systems that prevent it from doing that. We've not burned one pin yet. We've been working on this AI. We discuss wafer level burn-in for nine months.
We've not burned one pin, which is crazy. People are like, I don't understand how you do that. So there's some cool things about it that are just not intuitive that are as differentiated as anything, and we're excited about it. It's a fun market.

Operator

Larry Chlebina, Chlebina Capital.

Larry Chlebina

When was the first AI processor XP was that with the auto aligner? When was that up and running at that OSAT?

Gayn Erickson

That's a good question. We had it up and running here first. As those that have haven't visited us, they know we have these test cell labs, which we're upgrading. And we were actually shipping wafers out of our facility here. It's now up and running the first one, and we'll have the rest of them all completed this quarter, whatever, in the next few weeks here.

Larry Chlebina

But It was running at the OSAT, when?

Gayn Erickson

We haven't just always said it's up and running now.

Larry Chlebina

Chris, I got a quick question. What was the packaged part revenue in Q3?

Chris Siu

Package part in Q3, Well, in the past, we only disclose it in the year-end, but I would say, let me say it this -- I would say more than 20%.

Larry Chlebina

It's going to be in the Q but that all right. Gayn, is there any product launch on your flash customer that they're particularly wanting to use your systems on. I know this is an enterprise application. And I see these big SSDs coming up later this year that they have like 123 terabytes of data capacity on it, which if you do the math on it, they consume with a decent yield, each one consumes like wafer of NAND, just to stuff into that 2.5-inch form factor SSD.
Is there a product like that that's their interest in using your system to improve the yields because obviously, they're stacking a bunch of dies, is that the motivation?

Gayn Erickson

Okay. I would this is three in a row, we're answering a little bit indirectly. I just want to be a little careful of customers. But we talked to multiple. There's not that many NAND customers, and we have not talked to every single one of them, okay, for some geographical obvious reasons, okay?
But we've talked to several of them. And there is a general theme going on with those customers and with the industry if you watch what's going on. And that theme is -- so a couple of things.
In general, revenues are growing, okay? They were down last year and the year before, and they're growing again. But wafer capacity is not, okay? So there's not a lot of NAND fabs that are being coming online, and the capacity may be going to higher utilization, but the bits are exploding, right?
So what that means is that there's more and more density per wafer, okay? As you do that, the test methodology is either going to increase in test time or you're going to need to increase in power to address it?
The other thing that's going on is there's multiple kind of technologies out there that liked stacked or not stacked that's wrong one in, layers. So the NAND layers are going to hundreds of layers, which are just basically when I think of like a high-rise building, there's a lobby floor and then there's floors above. The lobby floor may have the logic on it for the interface and the floors above or what store all of the information, okay?
Now some people might say the I/O is on the top, but don't go into that. This is on one device. Then they take those devices and they stack them together into a package like you said with an SSD, okay? So one of the things that's going on in the memory itself is the die actually getting taller okay? And to hit the right things, the footprint, the XY footprint is getting smaller, which means the die count on a wafer is getting up.
So the way they're getting density is they're shrinking, they're getting the die, the footprint of that building to be smaller and they're making it taller, right? And that's how they're getting more out of a wafer, but that's causing problems with respect to parallelism and power, which is a key thing that would be driving our new requirement.
Now the other thing that's going on is this thing around hybrid bonding. And this is the first time I've talked about this on a call, okay? But on the hybrid bonding, there are several people out there that are introducing products that are coming up over the next year or so that what they do is they make that lobby floor, the logic IO on one process wafer and then they make all the buildings or the floors that hold all the data on another wafer.
And then they actually glue them together with a hybrid bond. When they do that, they can actually make the lobby floor smaller and faster because they can use a CMOS technology that will work without it, that will still -- that -- because they don't have to make a memory cell out of it. Then they take a different technology that is good at memory cells that is lousy at lobbies or the I/O speed and they can make those taller.
So now I can make the building taller by separating the NAND cells from the lobby floor I/O. And I just have to glue them together. And I say glue fancy. There are people that do this for living, but it's a hybrid bonding process that's similar to TSV bonding when you think of some of the other things that are going on in the world.

Larry Chlebina

If you do that for the flash -- for the NAND component of that, right?

Gayn Erickson

And/or the combined combination of it, et cetera. So there's some -- I don't want to get into too much of it, but there's technological things that are breaking the infrastructure. And those are the opportunities to get in because it's not that someone is growing in the new fab makes sense. And why would you want to go build a bunch of new equipment for that. So that's just -- that's the message I want to get across.
And that's provides an opportunity for things to be disruptive enough to actually change the way your process is doing. And it's not just one company. Ultimately, it would be all of them.

Larry Chlebina

I was just looking at the -- one of those SSD drives and the amount of NAND that they consume for a reasonable expectation on sales of those drives, you worked that number backwards and it seems like they would need something like $70 million worth of your equipment, just process those wafers in order to improve the yield that they get as they stack all those dies on top of each other into that small package.

Gayn Erickson

Yes. That was where I was going with in the prepared remarks that -- and I'm not trying to get -- there's a couple of different ways of cutting this thing. But if you're spending $80 billion -- if you're incrementally selling $80 billion a year of NAND and the process technology is changing to allow you to be competitive to get more bits to do whatever, what kind of spend would you be doing?
It's interesting if you're seeing that you're not necessarily spending it all on new capital equipment, your test budget might be bigger. So being able to spend $80 million, as you said, that's not at all out of reasons.

Larry Chlebina

It was pretty astounding that one customer, one flash memory customer on one product could require that kind of spend in order to get the yields where they have to get them. Anyway, I want to switch your fine pitch WaferPak that I guess you're developing for that application. Does that open up the DRAM high-bandwidth market for you guys?

Gayn Erickson

It opens it, yes, in terms of being able to test the DRAM. There's other things that would need to be done in order to do that, particularly around the DFT and low pin count test mode. But the interesting thing, and I've shared this, you've all people kind of track it over what the last -- how many quarters.
If you go back and look at what I said two years ago, I don't know whatever it was, just any quarter prior to AI, chat GPT, I kept saying that wafer line brand will first go to NAND. And then ultimately, it will go to DRAM. But in order for it to go to DRAM, the DRAM guys need to have the critical -- there needs to be more pain and the pain needs to be when they start stacking DRAM. And then when they do that, they'll have to figure out a low pin count mode and testability or us, they're going to get eaten up in yield.
So now what you see is with HBM, which is still not the biggest part, but it's certainly the fastest-growing part of DRAM. HBM requires this advanced packaging and stacking. And guess what, yields a big problem for them. And now you saw was it just last month, NVIDIA came out and told and said publicly, I'm now asking that all DRAM suppliers start supplying me with known good stacks of DRAM.
What does that mean? Well, obviously, it means they're not good. And I already told you 10 minutes ago that on my package part burn-in systems, when we burn in those devices, we're burning in the memories, too. Why are we burning in the memories there what a horrible place to do it. Anyhow, kind of interesting.

Larry Chlebina

But I thought the fine pitch WaferPak though, gets to the point where it can touch down on all those contact points on the DRAM. So opens that up. Am I missing something there?

Gayn Erickson

No, no, no, not really. Okay. So on the HPM memories, they may have 1,000 pins on them. They're all TSVs. They're at like 10 microns. No one can touch those TSV pads, not form factor, not anything.
There are still testability pads. But on a DRAM like that, maybe used to have 50 pins or 60 pens or something times 3,000 die. You go through that math, that's a big probe card. And so a full wafer probe card is quite expensive and cost effectiveness is it really there.
If you implement a DFT mode, Interestingly, like is done in the package, but anyhow, you can drop down to a JTAG I/O port or an MBIST I/O port, which is a single I/O pin. So why not do that that NAND guys did. So as that is implemented, you could see that people might drive towards the DRAM. And if any of the DRAM guys are listening, I'd love to talk to you again if I'm not already talking to you about this because I think we can help.

Larry Chlebina

They definitely have a yield issue. Last question real quick. There's a long-term customer of yours that's ramping up on a 3D packaging facility in New Mexico. And they're assembling three to five chiplets onto a sub chip. And I don't know, I think the throughput of that facility eventually will be at least $25 billion worth of product.
At what point do you go to that your customer and say you're already in that fab on the optical side? At what point do you say, hey, what would you spend if I could improve your yields, some of the things you said earlier, if I could improve your overall yield 3% on those heterogeneous chips, what would they spend to achieve that going forward?
I mean, every year, spend at once. Won't they spend like $300 million to get that capability. I know you'd be testing all the incoming component wafers before they get singulated and put it in the chiplets, but is that a hell of an opportunity for you guys?

Gayn Erickson

Okay. Well, I've never mentioned we have systems installed in New Mexico or all those things that you've said, and I won't confirm or deny anything here. I don't want to get myself in trouble with any potential customers. But someone who's doing that would certainly be attractive to us.
What we've definitely said and what you were implying there is that we have, what, six, seven optical customers related to silicon photonics, one in particular that has been our lead customer in a 10% or several years. And within those customers, we have worked with the largest to build this very high-power silicon photonics-based system that is amazing.
And we've qualified that. We shipped it for the first time last year. They spent the last year qualifying. We've been doing all kinds of engineering runs and things with different types of wafer packs. And we've heard that they are talking about ramping this year.
With these new optical devices, which the target for those are chip-to-chip or chiplet type things when we say that out loud. So I'm excited about that. I think as you said, well, why not use it in every other chip along the way. I think that's a good point. And if we can do very well with what we're planning right now, maybe that opens up an opportunity with our customer, if not others.
I would just as a reminder for everybody listening on the Incal, are tools set primarily on the wafer level until the InCal acquisition of package part, was a system that was designed around these megatrends with semiconductors that are growing to $1 trillion, where semiconductors are not more reliable year to year, they're less reliable due to line widths, the compound semiconductors like optical and silicon carbide stacking them all together in multi-chip modules or in modules like memories that are put together for SSDs or modules that are powered going to electric vehicles.
And then unfortunately, they're putting them in things that matter to the reliability like automotive devices or data centers or these very expensive AI processors. So that opens up an opportunity to do this burn-in stress conditions on those devices before they're put into the packages. And that's why the burn-in market now is so much bigger and more interesting than it has ever been in its history, at least going back to the early '80s because of it's enabling capabilities, technology otherwise you wouldn't have been able to.
We are picking up customers and markets along the way. And I always say [it a year] and we'll have another one. And Larry, if you have a specific contact on someone along those lines, it would not be the first time one of my shareholders has introduced me to a customer and I got an order out of it. So thank you.

Operator

Charles Tow, Private Investor.

Charles Tow

I think my first question has been asked by Larry. So my second question is, again, if everything is on track, when do you expect to receive mass production orders from the flash memory company, you are currently working with?

Gayn Erickson

Yes, it's a good question. What we've said right or wrong, so that we're trying to be relatively visible. It's always a little dangerous for all your potential competitors listening in is that this year, we spent the year working on some technological steps to prove the proof-of-concept of our low-cost test cell that is really around this wafer pack carrier that allows everything else to be lower cost and allows higher power, higher density.
If with that, our hope is upon successful completion of that, which we were targeting and trying to be around this next quarter or so, we then want to work with the customer around the next step, which would be in co-development of a test system that would go into that machine. We think that's going to take about a year or so. Maybe it could take longer, but there's reasonably maybe a year.
And so I wouldn't expect volume orders for that until the following year. So we're a year out or so, but you know what, if you told me I was taking volume orders and shipping for revenue in our fiscal 2027, which begins in June '26, I would tell you that's awesome.

Charles Tow

Yes, what size? How big the site you can imagine?

Gayn Erickson

How big is the site? The size of the market? Yes. I mean this is where -- what we think is if you were to look at -- and I don't want to pick on any customer, okay? But the market just this year is $80 billion.
I think it's growing at maybe 10%, 20% revenue. BIBs are growing 40%, something like that. So two years out from now, I mean, okay, maybe it's $100 billion, something like that. There, how much that is? You divide it by what there's six customers, five customers that matter.
They are not all even. But H1 is 20%, maybe a little less, okay. Then that spend by them is substantial and the cost savings advantage to be able to do that would be substantial. So in order for us to actually help them with yield, there's a considerable market.
This is embedded in the number that I said during my prepared remarks that the total available market is a combination of $500 million in test and $500 million in consumables. And the flash memory could be certainly 10%, if not more, of that overall market. So $100 million, $150 million, just wafer level burn-in for flash memory in that time frame minimum.

Operator

Okay. I show no further questions in the queue. I'd like to turn the floor back to management for any closing remarks.

Gayn Erickson

All right. Well, I appreciate everybody, as always, for jumping on here. I just want to -- before closing the call, just one personnel-related thing I hadn't gotten to. I'm just happy to announce the addition of a new member to the team, Didier Wimmers, to our organization as our EVP of Engineering. He's replacing Avi who as we previously announced, passed away unexpectedly last year.
Didier is an incredible addition to Aehr. He brings decades of direct experiences leading engineering teams in semi test. He's pretty well known. I joked around and said he's famous in the ATE space. He headed up all of Schlumberger ATE for many years, also managed the engineering team at FormFactor, who's basically one of the largest semiconductor test probe card companies in the world.
I work with FormFactor for many years in my memory experience before, and that's where I had met Didier in the past. So I've known him for a long time, maybe 20 years myself. He's a great leader, sense of knowledge in ATE, wafer probing, great experience. He's a great manager and also a nerdy guy, which I say very in a warm way.
He's a great member -- team member. I'm excited to have him on because of his experience in ATE system-level testing and even burn-in systems before joining Aehr. So we welcome him on board and excited to have him.
As always, we appreciate everybody joining in. If you happen to be anywhere near the Bay Area and are interested, we can try and set up something. Our facility here, we're getting right down to the final. We're about a month away from it being pretty well completed. And we've done a little bit of just a remodel in general for cubes and facility areas for the employees, which they really love.
But we did a major remodel in the manufacturing and the clean room labs which allow us to increase both manufacturing prototype engineering capacity, but particularly manufacturing capacity of our systems, our new packaged part systems from Sonoma that those systems will be coming over here over the next couple of months and also our WaferPaks and our BIM consumables from the package part side.
So all that will be consolidated here Q1-ish, I don't know, this summer, okay? But if you come by then, it will probably be finally happy with everything kind of put together and all the lights are up and everything, too.
So I thank you, everyone, for your encouragement. I've gotten a lot of calls of hanging in there. A lot of people very excited about the stuff that we're working on and kind of enduring through all these tariff things, but this too -- my favorite saying this too shall pass. So thank you all, and we'll talk to you next quarter.

Operator

This concludes today's conference, and you may disconnect your lines at this time. Thank you for your participation.